How candidates cheat in technical interviews and what are the consequences of this? (film, 6m)

There are two kinds of people in the world I don't trust. People who don't put swear words in their code comments, and people who don't tell the truth. With the rise of remote work and large language models over the last few years, we've also seen a rise in the number of people cheating on technical interviews, with an estimated 10% of people attempting to cheat. And today, we'll look at the actual techniques these clever wannabe programmers use to land a high-paying developer job in this highly competitive job market. And most importantly, the disastrous consequences for those who get caught. It is November 25th, 2024, and you're watching The Code Report. Some of my biggest inspirations growing up as a kid were athletes like Lance Armstrong, Barry Bonds, the entire Russian Olympic team, and Mike Tyson. They're all elite athletes, but they also bent the rules a little bit to get ahead. Now, I'm not recommending you bite your interviewer's ear off when you can't figure out how to invert a binary tree, but the reality of the modern tech job market is that many people out there are cheating to get ahead, and sometimes it actually works. But cheating happens in virtually every industry. Academics like to plagiarize, investors like to do Ponzi schemes, and chess players like to allegedly use remote-controlled Morse code-emitting implants that transmit stockfish moves directly to the sphincter. That's brilliant if true, but this technique wouldn't work very well on a programming interview. So now let's take a look at some real-world cheating techniques. The most basic technique is cribbing the answers. In high school, you may have written math equations in your palm before a big exam. In a remote technical interview, you can do the same thing by having a secret laptop out of the view of your webcam with a million browser tabs open to different LeetCode solutions. This works because many technical interview questions are reused over and over again. There are classics like FizzBuzz, Inverted Binary Tree, and A Magical String, and if you're lucky enough to get one of these questions, you can secretly look at your laptop to solve it. But solving the problem is only one part of the test. You also need to understand the logic behind the code. If you quickly solve a problem in code, but can't explain how it actually works, the interviewer will think one of two things. This guy is an autistic 10x developer unicorn, or more likely, you're just a filthy cheater. In a technical interview, it's crucial that you explain your thought process clearly as you solve a problem. You're expected to make errors and work towards a solution iteratively, just like you would in real life. But the most likely reason this cheating technique will fail is because you simply won't be asked a question on your secret laptop. A more sophisticated cheating technique is to study interview questions that have been leaked on the internet. Companies are always updating their interview questions, but that doesn't stop people from screenshotting or taking that content from the interview, then sharing it on various low-key websites and Discord channels. I'm not going to share these resources with you. Actually, no, f*** it. You can find leaked questions all over GitHub, as well as this weird Chinese website called 1.3acres.com. They compile things like the most common interview questions asked in the last six months at companies like Meta, but most of these are just public leak code questions anyway. But it becomes a gray area when proprietary non-standard questions become leaked. What doesn't constitute cheating, though, is discussing interviews with like-minded colleagues on awesome websites like daily.dev, the sponsor of today's video. It's a completely free social platform that curates all the best developer content on the internet and helps you connect with other like-minded people. Instead of scouring the bowels of Reddit for an update on your favorite JavaScript framework, you can rely on daily.dev to pull content from over a thousand top sources to get all the content you need in one place. And when you install their highly-rated browser extension, staying up-to-date becomes an easy daily habit. Join over one million other developers on daily.dev by using my invite link on the screen. But if you're not skilled enough to pass the technical interview, another way to cheat is to ask your bros for help. Pay a friend or two to sit next to you during the interview and deliver the solution to you in sign language. Typically, this needs to be done in person, because if you try to use some screen-sharing software during the interview, you'll likely get caught because many interview tools use pre-screen test to check to make sure that you're not using a remote desktop or some other collaboration tool. The main reason this won't work, though, is because most aspiring programmers don't have friends. But luckily nowadays, we have artificial programmers in the form of ChatGPT, Claude, and Gemini. Nowadays, you can simply screenshot a question, feed it into one of these AI tools, and get the solution immediately. But does this actually work? Earlier this year, Interview.io conducted a study to answer that question. They conducted interviews where the interviewers had no idea that the interviewees were using ChatGPT. They tested verbatim LeetCode questions, modified LeetCode questions, and completely custom questions. On verbatim questions, the pass rate was 73% for the cheaters, and only 53% on the non-cheating control group. On modified LeetCode questions, the pass rate went down to about 67%, still beating the control group. However, when it came to custom questions, the cheaters started to bomb and only passed at 25%. That's better than 0%, but the thing is, AI tools often spit out code that looks correct, but is actually nonsense. And if you try to pass that code off as legit in an interview, it'll be blatantly obvious that you're cheating. And that's exactly what happened to some guy who was trying to use Claude during an XAI interview. The interviewer called him out, then they had a nice discussion about how cheating actually occurs nowadays. One thing to keep in mind with this study, though, is that it was conducted about a year ago, and AI tools have already gotten much better at coding since then. And what's crazy is that in their experiment, nobody actually got caught in the act of cheating. Now, even if you manage to get away with cheating with AI during an interview, it's still a very bad idea, because if you interview with Fang, the next step is typically an in-person interview, where you'll have to problem-solve on a whiteboard. And your only hope of cheating there is with some kind of brain chip or remote-controlled suppository. But if you can pull that off, you deserve the job. But the main reason you shouldn't cheat is that you'll likely get caught. And when that happens, all kinds of bad things will happen. First of all, you'll be automatically rejected from that position. Second, you might even be blacklisted from reapplying to that company in the future. Third, if you're building a public profile on places like Twitter and LinkedIn, you'll find that tech is actually a pretty tight-knit community. And if people find out you're a cheater, it could permanently damage your reputation. Four, it'll be extremely embarrassing, and will be a bad decision that haunts you for the rest of your life. And five, in the unlikely case that you actually do get the job, you'll likely end up being a low performer who's at the top of the chopping block for the next round of layoffs. Ultimately though, cheating is not a mistake, it's a choice. So make sure to choose wisely. This has been The Code Report. Thanks for watching, and I will see you in the next one.

Menu

How candidates cheat in technical interviews and what are the consequences of this? (film, 6m)

Toggle timeline summary

Transcription